Table 5 Discriminative value AUC for the model including estrogen-related variants in the discovery and replication populations.

From: Steroid hormone-related polymorphisms associate with the development of bone erosions in rheumatoid arthritis and help to predict disease progression: Results from the REPAIR consortium

Discovery population (REPAIR Consortium; n = 816; RF += 571 and RF− = 238)

Replication population (DREAM registry; n = 436; RF += 328 and RF− = 94)

Demographic model (RF + Patients; n = 460)

LR test p-value

Demographic model (RF + Patients; n = 242)

LR test p-value

 

P-value

OR 95%CI

AUC 95%CIa

  

P-value

OR 95%CI

AUC 95%CIa

 

Gender (male)

0.004

0.486 (0.296–0.798)

  

Gender (male)

0.740

1.113 (0.590–2.100)

  

Age

0.00066

1.030 (1.013–1.048)

0.629 (0.567–0.692)

 

Age

0.654

0.995 (0.972–1.018)

0.528 (0.445–0.612)

 

Predictive model including 5 SNPs (RF + patients; n = 460)*

 

Predictive model including 4 SNPs (RF + patients; n = 242)Ϯ

 
 

P-value

OR 95%CI

AUC 95%CI a

  

P-value

OR 95%CI

AUC 95%CI a

 

ESR2 rs1271572

0.002

0.414 (0.236–0.726)

0.730 (0.672–0.780)‡

 

ESR2 rs1271572

0.763

0.899 (0.450–1.796)

0.625 (0.541–0.709)

 

CYP2C9 rs1799853

0.024

0.226 (0.062–0.824)

  

CYP2C9 rs1057910

0.058

3.385 (0.96–11.94)

  

CYP1B1 rs10012

0.013

0.442 (0.233–0.840)

  

CYP1B1 rs10012

0.014

0.285 (0.105–0.772)

  

CYP3A4 rs2740574

0.005

5.793 (1.718–19.53)

  

CYP3A4 rs2740574

0.075

0.370 (0.124–1.107)

  

SHBG r6259

0.011

2.316 (1.212–4.425)

  

SHBG r6259

0.444

1.378 (0.606–3.130)

  

Gender (male)

0.005

0.475 (0.281–0.803)

  

Gender (male)

0.941

0.976 (0.505–1.884)

  

Age

0.00039

1.033 (1.015–1.052)

 

2.46•10–8

Age

0.811

0.997 (0.972–1.022)

 

0.014

Demographic model (RF− patients; n = 182)

LR test p-value

Demographic model (RF− patients; n = 64)

LR test p-value

 

P-value

OR 95%CI

AUC 95%CI a

  

P-value

OR 95%CI

AUC 95%CI a

 

Gender (male)

0.713

0.864 (0.397–1.882)

  

Gender (male)

0.322

0.542 (0.161–1.820)

  

Age

0.053

0.978 (0.956–1.000)

0.588 (0.503–0.672)

 

Age

0.511

1.014 (0.973–1.056)

0.590 (0.429–0.752)

 

Predictive model including 5 SNPs (RF− patients; n = 182)

Predictive model including 4 SNPs (RF− patients; n = 64)Ϯ

 

P-value

OR 95%CI

AUC 95%CI a

  

P-value

OR 95%CI

AUC 95%CI a

 

ESR2 rs1271572

0.804

1.103 (0.509–2.388)

0.613 (0.530–0.696)

 

ESR2 rs1271572

0.828

1.214 (0.212–6.937)

0.778 (0.640–0.917)

 

CYP2C9 rs1799853

0.105

6.052 (0.688–53.26)

  

CYP2C9 rs1057910

0.219

0.287 (0.039–2.097)

  

CYP1B1 rs10012

0.521

1.526 (0.420–5.549)

  

CYP1B1 rs10012

0.493

2.316 (0.209–25.60)

  

CYP3A4 rs2740574

0.422

0.668 (0.250–1.785)

  

CYP3A4 rs2740574

0.033

0.113 (0.015–0.836)

  

SHBG r6259

0.699

0.870 (0.43–1.760)

  

SHBG r6259

0.029

0.139 (0.024–0.814)

  

Gender (male)

0.394

0.696 (0.303–1.600)

  

Gender (male)

0.287

0.453 (0.106–1.943)

  

Age

0.065

0.979 (0.956–1.001)

 

0.36

Age

0.949

0.998 (0.950–1.049)

 

0.015

  1. aIncluding age and gender as variables never dropped from models and when are compared with a baseline model with AUROC = 0.5. P ≤ 0.10 in bold (stepwise threshold).
  2. *All SNPs showing a significant association with erosive disease (P < 0.10) were initially added to the model in the discovery population.
  3. ‡A sort analysis in the discovery population revealed that this model showed an AUC value systematically higher than those observed in 50.000 randomized models:
  4. Average AUC of null distribution (50.000 models) = 0.644
  5. Z score = 6.79, PZ_score-value_(50.000perm) = 5.67•10−12.
  6. ϮAll SNPs were forced to be included in the replication population with the exception of the CYP2C9rs1057910 that was included due to the impossibility to calculate association estimates for the CYP2C9rs1799853 SNP.
Back to article page